Netskope, Inc.

Digital Guardian LLC

Niranjan Koduri

Emanoel Daryoush

Yi Zhang

Yihua Liao

Ari Azarafrooz

Jason B Bryslawskyj

Wayne Xin

Wai Tung Yim

Shreemathi Atreya

Richard Douglas LeCour

Image fingerprints (embeddings) are generated by an image fingerprinting model and indexed with an approximate nearest neighbors (ANN) model trained to identify the most similar fingerprint based on a subject embedding. For image matching, a score is provided that indicates a similarity between the input embedding and the most similar identified embedding, which allows for matching even when an image has been distorted, rotated, cropped, or otherwise modified. For image classification, the embeddings in the index are clustered and the clusters are labeled. Users can provide just a few images to add to the index as a labeled cluster. The ANN model returns a score and label of the most similar identified embedding for labeling the subject image if the score exceeds a threshold. As improvements are made to the image fingerprinting model, a converter model is trained to convert the original embeddings to be compatible with the new embeddings.

Image classification and accelerated classification training using deep learning image fingerprinting models and indexed embeddings

A system to generate an image classifier and test it nearly instantaneously is described herein. Image embeddings generated by an image fingerprinting model are indexed and an associated approximate nearest neighbors (ANN) model is generated. The embeddings in the index are clustered and the clusters are labeled. Users can provide just a few images to add to the index as a labeled cluster. The ANN model is trained to receive an image embedding as input and return a score and label of the most similar identified embedding. The label may be applied if the score exceeds a threshold value. The image classifier can be tested efficiently using Leave One Out Cross Validation ('LOOCV') to provide near-instantaneous quality indications of the image classifier to the user. Near-instantaneous indications of outliers in the provided images can also be provided to the user using a distance to the centroid calculation.

Image classifier automated testing and outlier detection

Image fingerprinting conversion between different image fingerprinting models

Image matching using deep learning image fingerprinting models and embeddings

Provided herein are systems and methods for classifying content to prevent data breach or exfiltration. An entity engine may receive content for classification into a content type for preventing data breach or exfiltration. The entity engine may determine that secondary data, defined by an operand of an entity definition, is present in the content. Each entity definition may correspond to one content type and may include a Boolean expression of operands. Each operand may include a matching element to be used for matching against content undergoing classification into one of the content types, upon secondary data defined by the operand being present in the content. The entity engine may classify the content into a content type of the content types, corresponding to the entity definition, based on matching the matching element of the operand to the content, and matching other operands of the entity definition to the content.

Systems and methods for identifying content types for data loss prevention

Provided herein are systems and methods for defining and securely sharing objects for use in preventing data breach or exfiltration. Memory may be configured to store a plurality of objects for use in preventing data breach or exfiltration. A validation engine can validate the objects, incorporate into each object an object identifier and a signature, and generate a subset of the objects for use by a first user. The validation engine can store, in the memory, the plurality of objects as a superset of objects corresponding to the generated subset. An evaluation engine may, responsive to identifying that one or more object identifiers and signatures in a received set of objects belong to the subset corresponding to the stored superset, verify whether any object in the received set has been tampered with.

Systems and methods for defining and securely sharing objects in preventing data breach or exfiltration

Growing community of inventors

Pleasanton, CA, United States of America

Niranjan Koduri